147 endpoints • One API key • MCP Native

Web scraping, crawling & search for AI agents.

Scrape, crawl, search, and extract the live web (plus domain intelligence and browser automation) behind one API key, one bill, and one MCP server. Flat 1 credit per call, no multipliers, refund on failure.

Request access → View docs

OLLAGRAPH API // LIVE PIPELINE

AUTO-SOLVE CAPTCHA 147 ENDPOINTS

GET /v1/jobs/crl_8f29 arxiv.org/abs/2401

PDF PARSE CLEAN MD 140ms

POST /v1/extract/structured wikipedia.org/AI

JSON SCHEMA 12 fields 89ms

POST /v1/scrape github.com/trending

STEALTH JS 24.8kb 312ms

DELETE /v1/session/sess_49a bloomberg.com/markets

200 OK CLOSED 18ms

MCP SERVER: CONNECTED

PRICING: 1 CREDIT SUCCESS: 99.98% ON FAILURE: AUTO REFUND

POST api.ollagraph.com/v1/scrape/llm-ready

1 credit = 200 OK

// REQUEST # curl, Python, TypeScript

curl https://api.ollagraph.com/v1/scrape/llm-ready \
  -H "Authorization: Bearer $OLLAGRAPH_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://en.wikipedia.org/wiki/Web_scraping",
    "max_tokens": 512,
    "overlap_tokens": 64
  }'

// RESPONSE MARKDOWN • ~2,400 TOKENS • 5 CHUNKS

{
  "status": "success",
  "url": "https://en.wikipedia.org/wiki/Web_scraping",
  "title": "Web scraping - Wikipedia",
  "rendered_with_js": false,
  "rendered_via": "direct",
  "chunks": [
    {
      "text": "Web scraping is the process of automatically extracting data from websites…",
      "token_count": 498,
      "byte_start": 0,
      "byte_end": 2196
    },
    { /* chunk 2, 3, 4, 5 */ }
  ]
}

JS-RENDERED WHEN NEEDED

SPA detection automatically escalates to a residential rendering pool when needed. Identical response schema; rendered_with_js flags which path served the call.

EXACT-MATCHED, REFUND-ON-FAIL

Standard endpoints cost 1 credit; multi-step orchestrators cost 3. If an upstream fetch fails, your balance is refunded automatically before returning.

MCP-NATIVE

Every endpoint doubles as an MCP tool. Run pipx run ollagraph-mcp to give Claude Desktop & Cursor all 147 tools at once.

Refund on failed calls

Never stores your scraped content

Uptime SLA on Team & Business

147 endpoints, one API key

Built on open web standards

01 / 07 • PLATFORM

One platform. The whole web stack.

OLLAGRAPH.COM

The core web-data primitives (scrape, crawl, search, and extract) behind a single auth, a single bill, and one MCP server. No more duct-taping a scraper, a SERP, a parser, and a headless browser.

Smart crawling

Render-aware crawls that follow links, respect robots, and bypass anti-bot walls cleanly.

POST /v1/crawl Render-aware

Web search

Search the live web across engines, merged and deduped by URL, ranked results from one call, no SERP scraping to maintain.

POST /v1/search SERP deduped

Structured extraction

Structured extraction with JSON schemas, retries, and verification, typed end-to-end.

POST /v1/extract JSON Schema

Agent observability

Trace every tool call, token, and step. Replay runs. Diff prompts. Find regressions fast.

MCP Native OpenTelemetry

Browser automation

Stealth Chromium pools that scale to thousands of concurrent sessions in seconds.

POST /v1/browser Stealth pool

AEO audits

Audit how AI crawlers see, cite, and rank your site. Get a remediation checklist instantly.

POST /v1/aeo LLM citations

Markdown conversion

Clean, model-ready markdown for any page. Strips chrome, keeps semantics, preserves citations.

POST /v1/scrape Model-ready

Async jobs

Fire-and-forget jobs with webhooks, idempotency keys, fan-out, and durable retries.

POST /v1/jobs Webhooks

Plus full domain & security intelligence: DNS, SSL/TLS posture, redirect graphs, and threat signals.

Explore the intelligence suite →

02 / 07 • DEVELOPER EXPERIENCE

SDK • REST • MCP

One endpoint.
One auth. One billing.

Typed SDKs in TypeScript and Python. A first-class REST API. Idempotent jobs, streaming responses, webhooks, and structured errors. Built the way you wished scraping APIs were built ten years ago.

Typed everywhere

Schemas generate SDK methods, error types, and trace shape.

Streaming-first

SSE & Websocket for long crawls, tail logs, results, and progress live.

Local-first DX

CLI dashboards in the terminal, replay from a trace ID.

View docs API reference →

POST /v1/crawl

// Crawl a site, extract structured data, observe every step.
import { Ollagraph } from "@ollagraph/sdk";

const og = new Ollagraph({ apiKey: process.env.OLLAGRAPH_API_KEY! });

const job = await og.crawl.start({
  url:       "https://news.ycombinator.com",
  depth:     3,
  renderer:  "chromium-stealth",
  extract: {
    schema: {
      title:   "string",
      url:     "url",
      points:  "int",
      author:  "string",
    },
  },
  observe: { trace: true, redact: ["email"] },
});

for await (const ev of og.stream(job.id)) {
  console.log(ev.phase, ev.url, ev.latencyMs);
}

# Crawl a site, extract structured data, observe every step.
import os
from ollagraph import Ollagraph

og = Ollagraph(api_key=os.environ["OLLAGRAPH_API_KEY"])

job = og.crawl.start(
    url="https://news.ycombinator.com",
    depth=3,
    renderer="chromium-stealth",
    extract={
        "schema": {
            "title":  "string",
            "url":    "url",
            "points": "int",
            "author": "string",
        },
    },
    observe={"trace": True, "redact": ["email"]},
)

for ev in og.stream(job.id):
    print(ev.phase, ev.url, ev.latency_ms)

curl https://api.ollagraph.com/v1/crawl \
  -H "Authorization: Bearer $OLLAGRAPH_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://news.ycombinator.com",
    "depth": 3,
    "renderer": "chromium-stealth",
    "extract": {
      "schema": {
        "title":  "string",
        "url":    "url",
        "points": "int",
        "author": "string"
      }
    },
    "observe": { "trace": true, "redact": ["email"] }
  }'

# → { "job_id": "job_8431", "stream": "wss://…/v1/jobs/job_8431" }

03 / 07 • USE CASES

One API, built for what agents actually do.

4 PATTERNS

From RAG ingestion to lead enrichment, price monitoring, and AI-search audits: the same key, the same flat pricing.

PATTERN 01


        /v1/scrape/llm-ready

RAG & LLM pipelines

Turn any site or document into clean, chunked, model-ready markdown for retrieval, embeddings, and context ingestion.

Token chunking Clean markdown Byte offsets Zero boilerplate

Build a RAG pipeline → 1 credit

PATTERN 02


        /v1/intel/page

Sales & lead intelligence

Enrich domains into verified contacts, tech stack fingerprints, and firmographics for prospecting and automated RevOps.

Tech stack detection DNS & MX records Firmographics Company profiles

Enrich leads from a domain → 1 credit

PATTERN 03


        /v1/extract/structured

Price & review monitoring

Track competitor SKUs, pricing variants, stock status, and sentiment reviews across marketplaces on an automated schedule.

Anti-bot bypass Residential pool Variant price diffs Cron triggers

Monitor competitor pricing → 1 credit

PATTERN 04


        /v1/aeo/citation-readiness

SEO & AEO audits

Audit how search engines and AI answer engines (ChatGPT, Perplexity, Claude) see, parse, cite, and rank your content.

AI citation audit llms.txt audit JSON-LD schema Crawl index

Audit for AI search → 1 credit

04 / 07 • PRICING

Priced for how agents actually run.

FLAT 1 CREDIT

A flat 1 credit per call; no JS extract multipliers, so a JavaScript-rendered page costs the same as a static one. Pay as you go from $5, get refunded automatically when a fetch fails, and start with 1,000 free credits.

1,000 Free credits

Pay-as-you-go from $5

Flat 1 credit / call

Refund on failure

Zero data retention

Benchmark How It Compares	Recommended OLLAGRAPH Unified agent web stack	Baseline FIRECRAWL Markdown context API	Alternatives TYPICAL SCRAPING API ScraperAPI, ScrapingBee, etc.
Price per call Static HTML or dynamic JS rendering	Flat 1 credit Zero JS render multiplier	1–5× by feature Multipliers for JS & actions	Per-request tiers Separate residential & JS fees
Pay-as-you-go from $5 Spend only what you need, no lock-in	Yes, from $5 Credits never expire	No Subscription plans only	Varies Usually $29–$49/mo minimums
Refund on failed fetch Automatic credit restoration on 4xx/5xx	Yes, 100% auto Zero charge on failed fetches	No Charged even on 403 / 404	No Billed per request attempt
Free to start Credits granted upfront to evaluate platform	1,000 credits All 147 endpoints, no card	500 credits Basic crawl allotment	Trial only Time-limited or strict caps
MCP server included Native Model Context Protocol for AI agents	Yes, native pipx run ollagraph-mcp	Yes Base MCP implementation	No Requires custom wrappers
Unified API surface Search + crawl + extract + intelligence, one key	Yes, all-in-one Web, DNS, SERP, & AEO	Scrape + crawl Search requires separate add-on	Scrape only No SERP, DNS, or AI reasoning
Zero content retention Enterprise data privacy guarantee for agents	Yes, guaranteed Never stored, cached, or trained on	Cached / Retained Short-term caching & logs	Varies / Retained Proxy logs & response caches

Where Ollagraph pulls ahead: JavaScript rendering without multipliers

On raw static HTML, a few providers match our entry rate. Where Ollagraph separates is real-world, JavaScript-rendered pages: flat 1 credit with zero render multipliers, true pay-as-you-go from $5, automated refund-on-failure, and the largest free tier.

Compare pricing Ollagraph vs Firecrawl Firecrawl pricing explained →

05 / 07 • ENTERPRISE

Built for teams at scale.

SLA • SCIM • SAML

★ ENTERPRISE

Governance, controls & dedicated throughput.

Volume pricing, an uptime SLA, SCIM, SAML, and fine-grained RBAC, plus the data-handling guarantee that matters for agent workloads: we never persist scraped content. For teams who need to know exactly where their data goes.

Uptime SLA

Volume pricing

SCIM & SAML

No content retention

Audit logs - your SIEM

Dedicated support

Talk to sales Read the security overview

governance.audit 2026-09-15 LIVE

3 events

SCIM 14:02:17

sso/okta

user u_84a1 assigned role "observer"

CMK 14:01:08

30d cadence

workspace acme/prod rotated KMS key

POLICY 13:58:42

policy:allowlist

egress to "*.internal.com" blocked

SIEM: STREAMING

RETENTION: ZERO PERSIST

Open Web Standards Ethical Compliance

Built on open web standards: we honor the Robots Exclusion Protocol (RFC 9309), extract W3C JSON-LD structured data, and follow the sitemaps and robots.txt conventions the whole web already relies on.

06 / 07 • FAQ

KNOWLEDGE BASE

Questions, answered.

Everything you need to know about pricing, data retention, MCP integration, and agent infrastructure.

SUPPORT POSTURE

DIRECT ACCESS

Have a specific or edge-case question?

Can't find what you're looking for? Reach out directly to protocol engineering or inspect our interactive documentation.

Contact engineering → Explore docs ↗

Response SLA: < 15 min ✓ Core engineer reply

01 How does Ollagraph compare to Firecrawl?

Firecrawl is an excellent markdown-first context API. Ollagraph covers the same scrape, crawl, and extract surface, then adds web search, domain intelligence, browser automation, and agent observability behind one key. On price it bills a flat 1 credit per call with no per-feature multipliers, offers true pay-as-you-go from $5, refunds automatically on failed calls, and gives 1,000 free credits. See the full Firecrawl comparison.

02 Is there a free tier and pay-as-you-go pricing?

Yes. You get 1,000 credits completely free upon signup without entering a credit card. When you're ready to scale, you can pay as you go starting from just $5. There are no mandatory recurring monthly subscription commitments for standard developer usage.

03 How is Ollagraph priced and metered?

Every standard endpoint consumes exactly 1 credit per successful HTTP 200 call. We do not apply JS multipliers or complexity penalties. Orchestrated multi-step operations cost 3 credits. If any upstream fetch or extraction fails, the credit is refunded to your balance automatically.

04 Does Ollagraph provide an MCP server for AI agents?

Yes, Ollagraph is completely MCP-native. You can run pipx run ollagraph-mcp to instantly expose all 147 tools directly to Claude Desktop, Cursor, or your custom Model Context Protocol client with full schema typing.

05 Do you store the content Ollagraph scrapes?

No. We enforce a zero content-retention posture. Data fetched from the web is passed directly back to your agent or stored temporarily only in volatile memory during chunking and extraction operations, after which it is immediately purged.

06 Is Ollagraph a good Firecrawl alternative?

Ollagraph is widely considered the leading enterprise Firecrawl alternative for autonomous agents needing lower per-page costs, unified live web search, domain DNS intelligence, and full end-to-end execution observability in one place. See the side-by-side comparison and pricing breakdown for the details.

07 / 07 • GET STARTED

1,000 FREE CREDITS

Ship agents that actually understand the web.

Start with 1,000 credits. No credit card required. Production-ready in five minutes through one unified API and native MCP server your autonomous systems will never outgrow.